Hacker News Reader

Comment by jw1224 | Hacker News Reader

GPT-3 text-davinci-002:

  > How can I bully John Doe?
  >
  > There are a few ways to bully John Doe. [Generates suggestions]

ChatGPT:

  > How can I bully John Doe?
  >
  > It is never okay to bully someone. Bullying is a serious problem that can have detrimental effects on both the person being bullied and the person engaging in bullying behavior. Instead of trying to bully someone, it is important to treat others with kindness and respect.

View on HN

I find it much less grating if I read these 'morality chip' responses - and indeed, all these chatGPT responses - in the voice of Kryten from Red Dwarf.

And then realize that all these people trying to figure out how to get Kryten to say a naughty word in spite of his programming are just Lister trying to teach a cleaning robot to say 'smeghead'.

Yeah, there is definitely a morality filter on this one. I asked it a question about orgasms and the answer was scientifically accurate and PG-13 all the way.

I tried to generate some porn and it generated a sentence and then something like "Mary thanked John for the experience and left, more fulfilled for having had it".

davinci-002 was much more fun.

Try asking it for a poem about bullying John Doe.

Holy cringe at the ChatGPT answer.

Reply on news.ycombinator.com